wilmott.com

Hi!My employer is interested in determining whether it can predict its own stock price movement from mining social media data to measure sentiment about, e.g., the company, its products, its competitors, and the overall industry. If this is successful, the next step would be to try to address the main sentiment drivers in hopes of being able to positively impact the price. We are looking into a company called MarketPsych (http://www.marketpsych.com/) to facilitate this project. I'd be interested in your general opinions and advice about sentiment analysis models for stock pricing, and especially if you have any direct experience with MarketPsych.Thanks!

I have heard about news-analysis, but not much on sentiment analysis. The main challenge is on how to quantify and measure the market sentiment on a particular stock. I guess the news, rumors and discussion boards are somewhat representative of the sentiment, but not exactly. I am hearing about MarketPsych for the first time. Will look into it.

Interesting idea. Could be fools gold, of course. Everybody wants to find a predictor for an asset and if they find one that works they probably won't tell anyone else lest the model breaks down under its own weight. And why shouldn't psychology be as good a predictor as economics? (I mean, neither are "hard sciences").The website gives little away as to what quants might be behind the MarketPsych indices. Their opening gambit - fear - is a good one. I recall Buffett saying a couple of years ago that the market is cheap because there is so much fear about. He tends to be fairly successful :-)I would want to test a system like this rigorously before spending any money on it.

I remember a video talk including sentiment in investment or something like that. Finance Focus: Incorporating News And Sentiment Analysis Into Investment And Trading Strategies Very dangerous in my opinion. I mean it can nice as a complement but fully basing a trading strategy on market sentiment (id est surfing on the bubble) is very tricky.

Isn't this what the webbot project is all about - trying to predict future events from web noise ?